On Fri, 2017-03-10 at 00:36 +0100, Bernd Schmidt wrote: > On 03/09/2017 08:28 PM, David Malcolm wrote: > > The root cause is an out-of-bounds memory write in the RTL dump > > reader when handling SYMBOL_REFs with SYMBOL_FLAG_HAS_BLOCK_INFO > > set. > > > > Such SYMBOL_REFs are normally created by > > varasm.c:create_block_symbol, > > which has: > > Hmm, I don't actually recall seeing this stuff. It's for section > anchors > apparently. > > > OK for trunk in stage 4? > > > > gcc/ChangeLog: > > PR bootstrap/79952 > > * read-rtl-function.c (function_reader::read_rtx_operand): > > Update > > x with result of extra_parsing_for_operand_code_0. > > (function_reader::extra_parsing_for_operand_code_0): Convert > > return type from void to rtx, returning x. When reading > > SYMBOL_REF with SYMBOL_FLAG_HAS_BLOCK_INFO, reallocate x to the > > larger size containing struct block_symbol. > > Looks OK for now, but longer term I think we should make it possible > to > reconstruct this data.
Thanks; fix committed to trunk as r246044. I'm also not very familiar with this part of RTL. print-rtl.c:rtx_writer::print_rtx_operand_code_0 has some special -casing for SYMBOL_REF, but if I'm reading things right we don't yet dump SYMBOL_REF_BLOCK and SYMBOL_REF_BLOCK_OFFSET, so we'd need to dump these somehow.
